Balance of Nanostructure and Bimetallic Interactions in Pt ModelFuel Cell Catalysts: In Situ XAS and DFT StudyDaniel Friebel,*,†,‡ Venkatasubramanian Viswanathan,§,∥ Daniel J. Miller,†,‡ Toyli Anniyev,†,‡

Hirohito Ogasawara,†,‡ Ask H. Larsen,⊥ Christopher P. O’Grady,§,∥ Jens K. Nørskov,§,∥

and Anders Nilsson*,†,‡,§,∥

†Stanford Institute for Materials and Energy Science, ‡Stanford Synchrotron Radiation Lightsource, and §SUNCAT Center forInterface Science and Catalysis, SLAC National Accelerator Laboratory, Menlo Park, California 94025, United States∥Department of Chemical Engineering, Stanford University, Stanford, California 94305, United States⊥Center for Atomic-Scale Materials Design, Department of Physics, Building 311, Technical University of Denmark, DK-2800Lyngby, Denmark

*S Supporting Information

ABSTRACT: We have studied the effect of nanostructuringin Pt monolayer model electrocatalysts on a Rh(111) single-crystal substrate on the adsorption strength of chemisorbedspecies. In situ high energy resolution fluorescence detectionX-ray absorption spectroscopy at the Pt L3 edge revealscharacteristic changes of the shape and intensity of the “white-line” due to chemisorption of atomic hydrogen (Had) at lowpotentials and oxygen-containing species (O/OHad) at highpotentials. On a uniform, two-dimensional Pt monolayergrown by Pt evaporation in ultrahigh vacuum, we observe asignificant destabilization of both Had and O/OHad due to strain and ligand effects induced by the underlying Rh(111) substrate.When Pt is deposited via a wet-chemical route, by contrast, three-dimensional Pt islands are formed. In this case, strain and Rhligand effects are balanced with higher local thickness of the Pt islands as well as higher defect density, shifting H and OHadsorption energies back toward pure Pt. Using density functional theory, we calculate O adsorption energies and correspondinglocal ORR activities for fcc 3-fold hollow sites with various local geometries that are present in the three-dimensional Pt islands.

1. INTRODUCTIONBimetallic catalyst materials are of great interest due to theirwide variability of the electronic structure that allows for“tuning” of the catalyst affinity to various reaction inter-mediates. Such a tuning is needed in particular for the oxygenreduction reaction (ORR) in fuel cells, where a significantreduction of Pt loading is essential for economic viability. Fortransition metal catalysts, the effect of bimetallic interactionscan be rationalized with a shift of the d-band center1−3 that isinduced by the lattice expansion or compression (strain effect)4

and by orbital interactions between the two different metals(ligand effect).4−6 The d-band shift corresponds to a change ofthe d occupancy and therefore directly affects the metal-adsorbate bond strength via the population of antibondingstates.3,4 Due to linear relationships between the d-band shiftand adsorption strength of various intermediates, the ORRactivity can be simply modeled by considering only oneintermediate, e.g., Oad, as “descriptor”, and a volcano-typerelation between adsorption strength and electrocatalyticactivity has been established using density functional theory(DFT).7,8 Nevertheless, it remains challenging to confirm theunderlying assumptions regarding structure and bonding at the

catalyst surface with in situ experiments in a condensedelectrolyte.Pt−O interactions can be probed via O 1s core-level

spectroscopies,9 but the low X-ray energies used in thesetechniques restrict the sample environment to ultrahighvacuum (UHV) or low-pressure (∼5 Torr) gas. By contrast,high-energy X-rays used in absorption spectroscopy at the Pt Ledges can readily penetrate liquid electrolytes.10−21 Recently,we demonstrated that high energy resolution fluorescencedetection X-ray absorption spectroscopy (HERFD XAS) at thePt L3 edge, applied to a well-defined Pt monolayer on aRh(111) substrate, is a surface sensitive probe of Pt−O surfaceinteractions that can unambiguously differentiate between thechemisorption of oxygen-containing species and surface oxideformation.22 Moreover, the nature of Pt oxides and theirformation mechanism at high potentials was found to bestrongly influenced by the underlying metal substrate.23

By systematically studying Pt monolayers on a variety ofM(111) substrates with HERFD XAS, we can directly probethe oxygen affinity of Pt “skin” layers under different strain and

Received: January 24, 2012Published: May 22, 2012

Article

pubs.acs.org/JACS

© 2012 American Chemical Society 9664 dx.doi.org/10.1021/ja3003765 | J. Am. Chem. Soc. 2012, 134, 9664−9671

ligand effects; this is an important experimental complement onadsorption enthalpy calculations used in the volcano plot. Incomparison of the latter with experimentally determined ORRactivities for a number of bimetallic systems, including Ptmonolayer catalysts,8 we note two important characteristics:first, with the exception of the more recent development ofPt3Sc and Pt3Y catalysts, the materials near the top of thevolcano are alloys of Pt with late 3d transition metals, which canbe unstable under fuel cell operating conditions due todissolution of the non-noble component. Second, it isremarkable that some model catalysts, i.e., Pt/Au(111), Pt/Ir(111), Pt/Rh(111), and Pt/Ru(0001),24 appear much moreactive than predicted theoretically. This raises the questionwhether an additional catalyst design criterion besides ligandand strain effects can alter the oxygen adsorption energy,enhancing the ORR activity for these systems.Here, we show that such an enhancement can be due to

three-dimensional nanostructuring of the Pt monolayer, whichhas not been accounted for in the previous studies, thusvalidating the theoretical predictions. Using the example of Pt/Rh(111), we compare electrochemical and in situ HERFD XASmeasurements in 0.01 M HClO4 electrolyte of two samplesobtained with different Pt deposition techniques resulting in auniform two-dimensional (2D) Pt layer and three-dimensional(3D) Pt islands, respectively. Most remarkably, we find almostcomplete absence of electrochemical and spectral signatures ofO/OHad and Had on 2D Pt/Rh(111), which directly confirmsthe significant down-shift of the Pt 5d-band that leads to 0.6 eVweaker O adsorption. The same vertical ligand and strain effectsalso weaken O and H adsorption on 3D Pt/Rh(111) but arepartially compensated in a large number of adsorption siteswith locally higher Pt thickness and under-coordinated Ptatoms at island edges and corners. We establish a linear scalingrelationship for the oxygen adsorption energy where local Ptthickness and under-coordination effects can be described in asingle parameter. For 3D Pt model islands whose structuralparameters are compatible with the extended X-ray absorptionfine structure (EXAFS) measured on 3D Pt/Rh(111), we find awidespread range of site-specific contributions to the ORRactivity whose average is in reasonable agreement with an RDEmeasurement on Pt/Rh(111).24 We predict that tailored 3DPt/Rh(111) nanostructures with increased occurrence of themost active sites could exhibit strongly enhanced ORR activity.

2. EXPERIMENTAL SECTION2.1. Sample Preparation. An 8 mm diameter commercial

Rh(111) single-crystal (Surface Preparation Laboratory, Zaandam,The Netherlands) was cleaned under ultrahigh vacuum by repeatedsputtering and annealing cycles. A home-built evaporator with resistiveheating was used to deposit the Pt film while the sample temperaturewas held at ∼600 K. The Pt coverage was monitored using changes inthe CO thermal desorption spectra from the Pt/Rh(111) surface.During the deposition, Pt atoms are incorporated into the topmost Rhlayer to form a surface alloy with increasing Pt content and, eventually,a pure two-dimensional Pt overlayer. In a previous study by Duisberget al., it was confirmed with ion scattering spectroscopy (ISS), whichstrictly probes only the first atomic layer, that the surface Pt contentreaches 100% before Pt atoms nucleate in a second layer.25

For the redox displacement of a Cu monolayer with Pt, the Rh(111)single crystal was prepared by flame annealing and subsequent coolingin a H2/N2 (5:95) atmosphere. The crystal was then transferred into anitrogen-filled glovebag where it was mounted in a hanging meniscusflow cell which allows for a rapid exchange of the electrolyte underpotential control. A single linear potential sweep was then started from+0.4 V (Ag/AgCl) in 1 mM CuSO4 + 0.05 M H2SO4 solution at a scan

rate of 10 mV/s in the cathodic direction, and the current peak arisingfrom Cu underpotential deposition26,27 (upd) was monitored.Immediately after the completion of the Cu upd peak, i.e., depositionof 1 ML Cu, the cell was flushed with the Cu-free supportingelectrolyte (0.05 M H2SO4). Meanwhile, the cathodic potential sweepwas continued to a final value of −0.4 V in order to prevent anydissolution of the Cu monolayer. In analogy to previous experi-ments28,29 which used the stabilization of a Cu upd layer on Au(111)in Cu-free electrolyte, we are able to verify with cyclic voltammetrythat the Cu layer remains stable and has the desired coverage of 1 ML.After the deposition of 1 ML Cu and its stabilization in Cu-freesupporting electrolyte, the Rh(111) sample was quickly transferred toa glass beaker where the polished surface was exposed to the Ptdeposition solution (1 mM K2PtCl4 + 0.05 M H2SO4) for ca. 30 s.

2.2. In Situ X-ray Electrochemical Cell. Subsequent to the Ptdeposition, the sample was mounted into an in situ X-rayelectrochemical cell. Two different setups have been used in thisstudy. A “thin layer” cell similar to setups which have been used earlierfor electrochemical surface X-ray diffraction studies30 was used with 6μm mylar foil (Goodfellow) as X-ray window. The cell can be deflatedto minimize the X-ray path length through the electrolyte solution. Adisadvantage of this setup is the limitation of mass transport due to thethin electrolyte layer. We circumvented this problem by limiting thepotential window to a range where the electrochemical currents underpotentiostatic conditions are low. Furthermore, for every change of thepotential, the cell was inflated in order to provide a thick electrolytelayer above the sample. At each new potential value, the cell was keptinflated for at least 15 min in order to allow for potential-inducedphase transitions to complete under thick-layer conditions beforereturning to the thin-layer mode for the next XAS measurement.Faster potential changes and higher electrochemical currents can beachieved in our “droplet” hanging meniscus cell,22 where a thickelectrolyte layer is provided throughout the X-ray measurement. Bothcells provide a three-electrode configuration with Pt wire as thecounter electrode and a leak-free Ag/AgCl reference electrode. Allelectrode potentials were converted to the scale of the reversiblehydrogen electrode (RHE).

All electrolytes were made from high-purity chemicals (70% HClO4,Trace Select Ultra, Sigma-Aldrich and 95% H2SO4, Trace Select,Sigma-Aldrich; 99.999% CuSO4·5H2O, Sigma-Aldrich) and ultrapurewater from a Millipore Gradient system.

2.3. X-ray Absorption Spectroscopy. All in situ X-rayabsorption spectra were measured at the Stanford SynchrotronRadiation Lightsource (SSRL). At Beamline 11-2, EXAFS measure-ments were carried out with conventional fluorescence detection usinga large 30-element Ge solid state detector. HERFD XANESmeasurements were performed at SSRL Beamline 6-2 using aSi(111) monochromator in combination with a Rowland circleanalyzer31 consisting of three spherically bent Ge perfect crystals (R= 1 m). The crystals were aligned in a backscattering geometry usingthe (660) Bragg reflection at 80.0° to select the Pt Lα1 fluorescenceline (9442 eV). The combined resolution of the monochromator andanalyzer as determined by measuring the elastic scattering was 1.6 eV.Assuming an intrinsic monochromator resolution of ∼1.3 eV theanalyzer resolution is estimated to be ∼1 eV.

For all XAS measurements, the incidence angle of the X-ray beam tothe Pt/Rh(111) surface was adjusted to the critical angle for totalexternal reflection, thereby enhancing the fluorescence intensity up to4-fold.32 The orientation of the electric field vector of the incidentbeam was perpendicular to the surface normal.

For the EXAFS data analysis, SIXPack33 was used for backgroundsubtraction, spline fitting, and least-squares fitting of the Fourier-transformed EXAFS signal. Backscattering phase and amplitudefunctions required for fitting of spectra were obtained from FEFF 6.34

2.4. Computational Methods. All HERFD XAS calculationswere carried out using the FEFF 8.4 program, which employs a fullmultiple-scattering formalism.35 By using the “NOHOLE” card,potentials and phase shifts were calculated assuming completescreening of the core-hole, resulting in better agreement withexperimental white-line intensities. This is fully consistent with

Journal of the American Chemical Society Article

dx.doi.org/10.1021/ja3003765 | J. Am. Chem. Soc. 2012, 134, 9664−96719665

previously reported FEFF results on transition metal L2 and L3edges.10,36−39 The line-sharpening effect observed in HERFD wasmodeled by reducing the theoretical lifetime broadening by 1.75 eVusing the “EXCHANGE” card. This value was determined bycomparing the computed XANES spectrum of a clean Pt/Rh(111)surface with an experimental spectrum at a potential corresponding tothe double-layer region, namely E = +0.4 V. More detailed informationabout the model structures and input parameters for the FEFF8calculations are provided in the Supporting Information, SI.2.5. DFT Calculations. All DFT calculations were performed using

GPAW40,41 within the ASE environment42 using the RPBE functionalfor exchange and correlation.43 The standard PAW setups and double-ζ polarized basis set provided with GPAW were used.44 We used a gridspacing of 0.18 Å, and a Brillouin zone sampling of 2×2 k-points alongthe periodic directions. O adsorption energies were calculated on smallPt clusters supported on the (111) fcc surface of Rh. The clusteradheres to the surface such that the Pt atoms all coincide withsubstrate lattice sites. The lattice constant was optimized with the Rhsubstrate using a separate DFT calculation. O binding energies werecalculated on each fcc hollow site on the (111) plane of the Pt cluster.No geometry optimization has been used. The O atom was alwayskept at a distance of 2.07 Å from the neighboring Pt atoms. The unitcell for the fcc surface of Rh contains 8 × 8 atoms in an orthorhombiccell and has five layers. A 5 Å vacuum is added between thenonperiodic cell boundary and the atom closest to that boundary.

3. RESULTS AND DISCUSSION

It is well-known that the fabrication of well-defined metalmonolayers can be challenging, especially in the case of Ptwhich in general, due to its high surface energy,45 is likely totend toward a Volmer−Weber growth mode rather than thedesired fully two-dimensional growth of one monolayer.However, on substrates with higher surface energies than Pt,such as Rh, Ru, or Ir,45 one would expect a growth mode ofeither the Frank−van der Merwe or Stranski−Krastanov type tobe favored, i.e., at least up to a coverage of 1 ML, Pt wouldgrow in a single 2D layer. The latter has been confirmed for thegrowth of Pt under UHV conditions on Rh(111)25 andRu(0001).46 While a well-defined 2D Pt monolayer wassuccessfully prepared in ultrahigh vacuum (UHV) by Ptvapor deposition, we discovered that 3D island growth occurswhen an electrochemical preparation is chosen, which consistsof the redox displacement of an underpotential deposited (upd)Cu monolayer.24,47

The morphology of the deposited Pt layers on Rh(111) wasdetermined for both UHV and electrochemically preparedsamples using in situ Pt L3 extended X-ray absorption finestructure (EXAFS), recorded at potentials close to hydrogenevolution. The Fourier transformed EXAFS magnitudes forboth samples are shown in Figure 1a,b. Least-squares fittingwith Pt−Pt and Pt−Rh nearest-neighbor coordination shellsgives coordination numbers (Table 1) that can be used todetermine the film morphology. For the vapor depositedsample, we found very good agreement with a pseudomorphicPt layer of 1 ML thickness which uniformly covers the Rhsurface. In contrast, the significantly smaller Pt−Rh coordina-tion number in the redox-displacement sample indicates thethree-dimensional nature of the deposit where only ∼50% ofthe Pt atoms are in direct contact with the Rh substrate. Sincethe Pt−Pt coordination number, at the same time, is alsosignificantly below values that would be expected for a uniformbilayer or multilayers, there must be a large number of under-coordinated Pt atoms. The observed coordination numbers canbe explained with a model structure consisting of three-dimensional islands. After a detailed consideration of various

island model structures (SI) we find agreement with theEXAFS results for 3D Pt/Rh(111) within experimental errorbars for islands of which the most range from two to four layersthickness and ∼1 to ∼4 nm lateral width. This range of islandwidths is in very good agreement with the sizes of Pt islandsthat can be seen in an in situ STM measurement47 of Pt/Rh(111), where the same redox-displacement technique wasemployed.The near-edge region (XANES) of Pt L3 spectra shows a

characteristic absorption maximum (“white-line”) due to 2p→5d transitions and thus provides a probe of the unoccupied partof the Pt 5d-band.48,49 The information about the energydistribution of unoccupied d states is limited by the Pt 2p corehole lifetime broadening, but significantly sharpened spectralfeatures can be obtained in the High Energy ResolutionFluorescence Detection (HERFD) mode36,50 which we used inour experiment. The HERFD technique, together with the useof well-defined single-crystal samples and sufficiently largemodel structures in the multiple-scattering computations,eliminates uncertainties in the interpretation of in situ XAS,in particular at high electrochemical potentials where contra-dictory models of Pt−O interactions have been proposed.11−20

HERFD XAS spectra of both samples in N2-saturated 0.01 MHClO4 (Figure 2) were recorded in situ in order of increasing

Figure 1. Morphology determination for Pt/Rh(111) using EXAFS.Fourier transforms of measured in situ EXAFS, fitting results for (a)UHV evaporated Pt/Rh(111), (b) Pt/Rh(111) prepared by redoxdisplacement of a Cu upd layer. Bottom of (a) and (b): schematicstructure models.

Table 1. In Situ EXAFS Fitting Results

Pt−Pt Pt−Rh R factor

2D Pt/Rh(111)a 0.0282N 6.5 ± 0.8 3.2 ± 0.7R (Å) 2.72 ± 0.02 2.72 ± 0.02σ2 (Å2) 0.005 0.0053D Pt/Rh(111)b 0.0238N 7.3 ± 0.7 1.4 ± 0.5R (Å) 2.74 ± 0.02 2.68 ± 0.05σ2 (Å2) 0.005 0.005

aData range: k = 3.0−9.8 Å−1. bData range: k = 3.0−9.5 Å−1.

Journal of the American Chemical Society Article

dx.doi.org/10.1021/ja3003765 | J. Am. Chem. Soc. 2012, 134, 9664−96719666

potential with respect to the reversible hydrogen electrode(RHE). After increasing the potential to 1.1 V and highervalues, both samples consistently show a strong broadening andintensity increase of the white-line and intensity decrease atenergies just above the white-line region, which we identifiedearlier as an unambiguous signature of Pt oxide formation.22

Besides, the two samples show significant differences, both inthe overall white-line intensities that are generally larger for the2D Pt layer (Figure 2a), as well as in the potential-dependentchanges of white-line shape and intensity. Additional changescan be seen for the 3D Pt islands (Figure 2b) in two potentialregions. At potentials close to hydrogen evolution, a reductionof the peak intensity together with a significant shoulder on thehigh energy side (∼11570 eV) of the white-line is caused bychemisorbed H. A subtle increase of the peak intensity at 11566eV occurs at potentials just below 1.1 V, which is due to O orOH adsorption. As shown in Figure 3a, this interpretation ofthe spectral changes can be qualitatively confirmed with abinitio multiple-scattering calculations on model structures of H/

Pt/Rh(111) and O/Pt/Rh(111) using the FEFF8 code.35 It isimportant to note that O chemisorption causes only an increaseof the peak at 11566 eV with no significant broadening. This isin strong contrast to the spectral signatures of Pt oxides (Figure3b), where the white-line has much higher integrated intensityand is shifted toward higher energy.The assignment of the Pt−H signature is also in agreement

with previous studies where Pt nanoparticles were used in themeasurements, although the corresponding FEFF calculationswere employing a very small Pt6 cluster.

21

In order to represent the measured spectral changesquantitatively, we performed a deconvolution using anarctangent function to model the absorption edge and twoGaussian profiles to represent the white-line peak, respectively.We kept the energy position of each component and the widthand step height of the arctangent function at fixed values andallowed only the widths and intensities of the Gaussianfunctions to vary with the potential. For all spectra, good fittingresults with reduced χ2 values better than 0.0006 could thus beobtained. The results for both samples including theiruncertainties are shown in Figure 4. The white-line for bothsamples at potentials below 1.1 V can be fitted with a sharp“main” peak at ∼11 566 eV and a broad peak at 11 568.4 eVrepresenting the high-energy shoulder. At 1.1 V, Pt oxideformation causes the higher energy component to becomesignificantly sharper and more intense, and upon furtherpotential increase it eventually exceeds the intensity of thecomponent at lower energy.Two additional phase transitions with more subtle spectral

features appear on the 3D Pt islands: at low potentials, thesignature of chemisorbed hydrogen can be clearly seen in thedepressed peak at 11 566 eV and stronger peak at 11 568 eV,and a subtle increase of the component at 11 566 eV as thepotential is raised from 0.8 to 1.13 V indicates O/OHchemisorption. Both of these changes are virtually absent forthe 2D Pt sample. The suppression of features induced by Pt−H and Pt−O can only be explained when we assumesignificantly lower adsorbate coverages on the 2D Pt layer.This interpretation is further supported by the comparison of

cyclic voltammetry (CV) curves for 2D Pt/Rh(111) and 3DPt/Rh(111) (Figure 5). In the CV of 2D Pt/Rh(111), weobserve that both onsets of Had and O/OHad formation aresignificantly shifted outward with respect to the potentialwindow between 0 and 1.0 V, resulting in strongly reducedadsorbate coverages as compared to Pt(111). By contrast, theCV of 3D Pt/Rh(111) has much more pronounced featurestoward both ends of the potential window. The broad shape ofthe anodic current increase due to O/OHad formation indicatesthe presence of (100) or (110) facets, further supporting theproposed 3D island model. The large width of the potentialrange of O/OH adsorption can also explain why thecorresponding changes in HERFD XAS are more difficult todetect than the spectral signature of Had.In order to further elucidate the effect of nanostructure and

local coverage on the shape of the CVs, we prepared anadditional Pt/Rh(111) sample in UHV with a coverage of 3 MLPt. In this case, also shown in Figure 5, the inward shift of bothHad and O/OHad features at the cost of the double layer rangecan be clearly seen, without significant additional features fromthe contribution of (100) or (110) facets. Even for 3 ML Pt/Rh(111), however, the double layer range with an adsorbate-free or H2O-covered Pt surface is wider than for purePt(111);51 this can be attributed to the strain effect.

Figure 2. In situ Pt L3 HERFD XAS for 1 ML Pt/Rh(111) in 0.01 MHClO4: (a) 2D Pt film, (b) 3D Pt islands. Spectra were recorded inthe order of increasing electrochemical potentials.

Figure 3. Calculated HERFD XAS, using the FEFF8 code, for clean,Had- and Oad-covered Pt/Rh(111).

Journal of the American Chemical Society Article

dx.doi.org/10.1021/ja3003765 | J. Am. Chem. Soc. 2012, 134, 9664−96719667

Our voltammetry results for UHV-prepared Pt/Rh(111)samples with different Pt layer thickness show exactly the sametrend as a combined UHV-STM and voltammetry study ofsimilar Pt layers where a Ru(0001) substrate was used.46 Bothof these experimental findings confirm the DFT-basedprediction that 1 ML Pt on Rh(111) and Ru(0001) adsorbO ∼0.6 and ∼0.7 eV more weakly than Pt(111), respectively.8

These adsorption energy changes correspond to positive shiftsof the equilibrium potential for Oad of ∼0.3 and ∼0.35 V for 1ML Pt on Rh(111) and Ru(0001), respectively. Approximatelythe same potential shifts can be expected for OHad due to thescaling relationship between O and OH adsorption energies.52

In order to estimate O adsorption energies on 3D Pt/Rh(111), extensive DFT calculations were carried out using 12model structures (Figure 6) representing different local Ptcoverage, as well as different proximity of the O fcc 3-foldhollow adsorption site to surrounding Pt and Rh atoms and

vacancy sites. Two effects can be seen: the weakening of the Oadsorption energy compared to pure Pt(111), ΔEO, isattenuated by ∼0.1 eV when we increase the local Pt thicknessfrom 1 to 2 ML, and O adsorption is further enhanced when wemove the fcc 3-fold hollow adsorption site from the center ofthe island to the proximity of an edge with under-coordinatedPt atoms. To disentangle ligand, strain, and local under-coordination effects, we carried out a simple linear regressionanalysis to correlate the O binding energy with the averagecoordination numbers NPt

S and NRhS of Pt and Rh surrounding

the three Pt atoms that constitute the adsorption site. Wefound the binding energy of O to scale linearly with an effectivecoordination number given by Neff

S = NPtS + 1.09 NRh

S. The

Figure 4. Deconvolution of in situ HERFD XAS of Pt/Rh(111) in0.01 M HClO4. While only the Pt metal-to-oxide transition can beidentified for the 2D Pt layer (a), the 3D deposit (b) shows additionalspectral signatures due to Had and O/OHad. Three representativefitting results for 3D Pt/Rh(111) are shown in (c) using the samecolors as in (b) for the peak areas, and a gray line for the arctangentfunction.

Figure 5. Cyclic voltammograms of UHV-evaporated Pt/Rh(111)samples with two different thicknesses (2D Pt/Rh(111), ΘPt = 1 ML,and ΘPt = 3 ML), and 3D Pt/Rh(111).

Figure 6. DFT calculations of the adsorption energy ΔEO of a single Oatom on Pt/Rh(111) model structures, shown as relative valuescompared to Pt(111) (ΔEO,Pt(111) = 0). The corresponding adsorptionenergy for each structure is plotted as a function of the effectivecoordination number around the O adsorption site, and a linear fit(red line) is shown. The linear relationship can be used to estimateΔEO for any other fcc 3-fold hollow site on Pt islands with variousshapes and sizes.

Journal of the American Chemical Society Article

dx.doi.org/10.1021/ja3003765 | J. Am. Chem. Soc. 2012, 134, 9664−96719668

coefficient for NRhS represents the ligand effect which in the

case of Rh weakens the metal−oxygen bond more strongly thanPt.This relation also serves as a guiding principle to determine

the electrocatalytic activity for different morphologies ofnanostructures as it gives a simple way to correlate the oxygenbinding energy to an easily calculable geometric parameter.Although this linear relationship is surprisingly simple, itsvalidity becomes evident when we consider that (i) it predictsapproximately the same gain in adsorption energy near islandedges as an earlier model for O adsorption near step edges onPt(111);53 (ii) it predicts no significant ligand effect beyond 1ML local Pt thickness; this result is similar to the calculation ofH adsorption energies on Pt/Ru(0001) where the ligand effectbecomes very small already for 2 ML Pt.46

The linear correlation in Figure 6 can be applied to allpossible fcc 3-fold hollow sites (SI) that can occur within the3D Pt island morphology determined with EXAFS. Thecontribution of each individual site to the overall ORR activitycan be estimated using the volcano relationship8 with ΔEO, anddirectly compared with the previous RDE measurement on Pt/Rh(111) (Figure 7a).24 Although we can only speculate aboutthe precise size and shape distribution of 3D Pt islands in oursample and the corresponding occurrence of adsorption sites, itappears that the average of site-specific ORR activitycontribution is reasonably close to the experimental value.We propose that a kinetically controlled 3D growth mode

can be generalized for Pt/M(111) (M = Au, Pd, Rh, Ir) and Pt/Ru(0001) monolayer catalysts prepared by the redox-displace-ment technique;24 this explains why their measured ORRactivities, which are also shown in Figure 7a, deviate from theprediction, where a 2D Pt layer was assumed,8 toward pure Pt.In the Pt/Au(111) system, potential-dependent Pt/Au placeexchange reactions54 could further add complexity.It is remarkable that the contribution of under-coordinated

sites exceeds that of removing the vertical ligand effect and caneven shift ΔEO for individual adsorption sites almost exactly to0.20 eV, where the volcano reaches its maximum.8 Since suchan “optimal” adsorption site near the top corners of islands is inclose proximity to a more strongly adsorbing bridge or 4-foldhollow site, we assume that it cannot be occupied due torepulsive O−O interactions. Nevertheless, we can identify twoother types of fcc 3-fold hollow sites that could offer a ∼9.5-foldactivity enhancement over a midterrace fcc site on Pt(111).These two most active sites can be found along island edges ofthe “B” type which separate two adjacent (111) facets. Theiroccurrence could be maximized through shape-selectivenanofabrication of bilayer Pt islands as shown with arepresentative example in Figure 7b. We also show predictedORR activities for a series of bilayer islands of analogous shapebut with different lateral widths; a detailed description of theunderlying analysis is given in the SI. Within the series in Figure7b, only midterrace sites and the two most active types of near-B-edge sites contribute to the ORR activity. Since the numberof midterrace sites grows faster than that of near-edge sites withincreasing island width, the smallest islands exhibit thestrongest ORR enhancement of up to 5-fold over Pt(111).Such high ORR activities have been reached with Pt3Co andPt3Ni alloys,55 but a Pt/Rh catalyst offers much higherresistance against degradation due to the combination of twometals with high cohesion energy and stability againstdissolution. While the use of Rh as catalyst component isunlikely to be cost-efficient, the same design principle could be

used for other bimetallic combinations of Pt with moreabundant materials, exploiting the advantage of catalyst stabilitythrough high cohesion energy, and at the same timecompensating a d-band shift toward weaker O adsorption bygenerating tailored nanostructures with under-coordinated Ptsites.

5. CONCLUSIONSDue to the design of the model catalyst and the reducedlifetime broadening in the HERFD mode employed here, our insitu XAS measurements on Pt/Rh(111) monolayer electro-catalysts reveal the subtle spectral features of chemisorbed O/OH more clearly than conventional XAS. Thus, the previousuncertainties in distinguishing O/OH from Pt surface oxidehave been overcome. Moreover, our experimental results clearlyand consistently show significant differences in surfacemorphology and chemisorption behavior as a result of thetwo sample preparation techniques employed here.

Figure 7. (a) Predicted individual ORR activities of fcc 3-fold hollowsites on 3D Pt/Rh(111). For comparison, measured ORR activities areshown for a range of bimetallic electrocatalysts5,8,24 including theprevious measurement on 3D Pt/Rh(111).24 The latter can beconsidered to represent a weighted average of the individual sitecontributions plotted with light green, dark green and yellow markers.(b) Bilayer islands with maximized occurrence of near-B-edge sites(yellow). The graph shows the ORR activity enhancement for a seriesof such islands whose lateral width is varied by changing the length ofthe B-edges while the A-edges are fixed at 0.54 nm length.

Journal of the American Chemical Society Article

dx.doi.org/10.1021/ja3003765 | J. Am. Chem. Soc. 2012, 134, 9664−96719669

DFT calculations have predicted before and confirm herethat atomic oxygen is bound to a Pt monolayer supported onRh(111) ∼0.6 eV more weakly than to the surface of pure Pt;8

this is due to a shift of the d-band center to lower energy causedby strain and vertical ligand effects. Three-dimensional Ptislands on Rh(111) show significantly higher affinity tochemisorbed H at low potentials and O/OH at high potentialsthan the two-dimensional monolayer on the same substrate.This can be explained with two cooperative effects that bothcause an upshift of the Pt 5d-band and a corresponding increaseof adsorption enthalpies for both H and O/OH. The higherlocal thickness (2−4 ML as determined from EXAFS) in 3DPt/Rh(111) removes to a great extent the vertical ligand effect,leaving only the strain effect. Hence, the d-band center isshifted back from the calculated position for 2D Pt/Rh(111)toward that for pure Pt, resulting in an increase of both Pt−Oand Pt−H bond strengths.The high amount of edges and corners in the 3D island

nanostructure adds further enhancements in O affinity in a verysimilar manner to stepped Pt(111) surfaces.53 This canovercompensate ligand and strain effects, thus opening a newfield of catalyst design through nanostructuring, where catalyticactivity becomes much less dependent from the choice ofconstituent elements. Hence, the often-encountered incompat-ibility between catalyst activity and stability could be overcome.

■ ASSOCIATED CONTENT*S Supporting InformationDetailed consideration of model structures for the EXAFSinterpretation; details of FEFF8 calculations; calculated ORRactivities for various Pt islands on Rh(111). This material isavailable free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATIONCorresponding Authordfriebel@slac.stanford.edu; nilsson@slac.stanford.edu

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThis work is supported by the Department of Energy, Office ofBasic Energy Sciences, Division of Materials Sciences andEngineering, under contract DE-AC02-76SF00515 and Divi-sion of Chemical Sciences through the SUNCAT Center forInterface Science and Catalysis. This research was partly carriedout at the Stanford Synchrotron Radiation Lightsource, aNational User Facility operated by Stanford University onbehalf of the U.S. Department of Energy, Office of Basic EnergySciences. D.F. is grateful to the Alexander von HumboldtFoundation for a Feodor Lynen Fellowship. The Center forAtomic-Scale Materials Design is funded by the LundbeckFoundation. We acknowledge support from the Danish Centerfor Scientific Computing.

■ REFERENCES(1) Hammer, B.; Nørskov, J. K. Nature 1995, 376, 238−240.(2) Hammer, B.; Nørskov, J. K. Surf. Sci. 1995, 343, 211−220.(3) Nilsson, A.; Pettersson, L. G. M.; Hammer, B.; Bligaard, T.;Christensen, C. H.; Nørskov, J. K. Catal. Lett. 2005, 100, 111−114.(4) Strasser, P.; Koh, S.; Anniyev, T.; Greeley, J.; More, K.; Yu, C.;Liu, Z.; Kaya, S.; Nordlund, D.; Ogasawara, H.; Toney, M. F.; Nilsson,A. Nat. Chem. 2010, 2, 454−460.

(5) Stamenkovic, V. R.; Fowler, B.; Mun, B. S.; Wang, G. F.; Ross, P.N.; Lucas, C. A.; Markovic, N. M. Science 2007, 315, 493−497.(6) Wang, C.; Chi, M.; Li, D.; Strmcnik, D.; van der Vliet, D.; Wang,G.; Komanicky, V.; Chang, K.-C.; Paulikas, A. P.; Tripkovic, D.;Pearson, J.; More, K. L.; Markovic, N. M.; Stamenkovic, V. R. J. Am.Chem. Soc. 2011, 133, 14396−14403.(7) Nørskov, J. K.; Rossmeisl, J.; Logadottir, A.; Lindqvist, L.;Kitchin, J. R.; Bligaard, T.; Jonsson, H. J. Phys. Chem. B 2004, 108,17886−17892.(8) Greeley, J.; Stephens, I. E. L.; Bondarenko, A. S.; Johansson, T.P.; Hansen, H. A.; Jaramillo, T. F.; Rossmeisl, J.; Chorkendorff, I.;Nørskov, J. K. Nat. Chem. 2009, 1, 552−556.(9) Miller, D.; Oberg, H.; Kaya, S.; Sanchez Casalongue, H.; Friebel,D.; Anniyev, T.; Ogasawara, H.; Bluhm, H.; Pettersson, L. G. M.;Nilsson, A. Phys. Rev. Lett. 2011, 107, 195502.(10) Teliska, M.; Murthi, V. S.; Mukerjee, S.; Ramaker, D. E. J.Electrochem. Soc. 2005, 152, A2159−A2169.(11) Vukmirovic, M. B.; Zhang, J.; Sasaki, K.; Nilekar, A. U.; Uribe,F.; Mavrikakis, M.; Adzic, R. R. Electrochim. Acta 2007, 52, 2257−2263.(12) Zhang, J.; Mo, Y.; Vukmirovic, M. B.; Klie, R.; Sasaki, K.; Adzic,R. R. J. Phys. Chem. B 2004, 108, 10955−10964.(13) Zhang, J.; Sasaki, K.; Sutter, E.; Adzic, R. R. Science 2007, 315,220−222.(14) Teliska, A.; O’Grady, W. E.; Ramaker, D. E. J. Phys. Chem. B2005, 109, 8076−8084.(15) Mukerjee, S.; Srinivasan, S.; Soriaga, M. P.; Mcbreen, J. J.Electrochem. Soc. 1995, 142, 1409−1422.(16) Mukerjee, S.; McBreen, J. J. Electroanal. Chem. 1998, 448, 163−171.(17) Tada, M.; Murata, S.; Asakoka, T.; Hiroshima, K.; Okumura, K.;Tanida, H.; Uruga, T.; Nakanishi, H.; Matsumoto, S.; Inada, Y.;Nomura, M.; Iwasawa, Y. Angew. Chem., Int. Ed. 2007, 46, 4310−4315.(18) Imai, H.; Izumi, K.; Matsumoto, M.; Kubo, Y.; Kato, K.; Imai, Y.J. Am. Chem. Soc. 2009, 131, 6293−6300.(19) Allen, P. G.; Conradson, S. D.; Wilson, M. S.; Gottesfeld, S.;Raistrick, I. D.; Valerio, J.; Lovato, M. J. Electroanal. Chem. 1995, 384,99−103.(20) Principi, E.; Witkowska, A.; Dsoke, S.; Marassi, R.; Cicco, A. D.Phys. Chem. Chem. Phys. 2009, 11, 9987−9995.(21) Teliska, M.; O’Grady, W. E.; Ramaker, D. E. J. Phys. Chem. B2004, 108, 2333−2344.(22) Friebel, D.; Miller, D. J.; O’Grady, C. P.; Anniyev, T.; Bargar, J.;Bergmann, U.; Ogasawara, H.; Wikfeldt, K. T.; Pettersson, L. G. M.;Nilsson, A. Phys. Chem. Chem. Phys. 2011, 13, 262−266.(23) Friebel, D.; Miller, D. J.; Nordlund, D.; Ogasawara, H.; Nilsson,A. Angew. Chem., Int. Ed. 2011, 50, 10190−10192; Angew. Chem. 2011,123, 10372−10374.(24) Zhang, J. L.; Vukmirovic, M. B.; Xu, Y.; Mavrikakis, M.; Adzic,R. R. Angew. Chem., Int. Ed. 2005, 44, 2132−2135.(25) Duisberg, M.; Drager, M.; Wandelt, K.; Gruber, E. L. D.;Schmid, M.; Varga, P. Surf. Sci. 1999, 433, 554−558.(26) Wu, Z.-L.; Zang, Z.-H.; Yau, S.-L. Langmuir 2000, 16, 3522−3528.(27) Anjos, D. M.; Rigsby, M. A.; Wieckowski, A. J. Electroanal. Chem.2010, 639, 8−14.(28) Friebel, D.; Schlaup, C.; Broekmann, P.; Wandelt, K. Surf. Sci.2006, 600, 2800−2809.(29) Friebel, D.; Schlaup, C.; Broekmann, P.; Wandelt, K. Phys.Chem. Chem. Phys. 2007, 9, 2142−2145.(30) Ocko, B. M.; Wang, J.; Davenport, A.; Isaacs, H. Phys. Rev. Lett.1990, 65, 1466−1469.(31) Glatzel, P.; Bergmann, U. Coord. Chem. Rev. 2005, 249, 65−95.(32) Waychunas, G. A. Rev. Mineral. Geochem. 2002, 49, 267−315.(33) Webb, S. M. Phys. Scr. 2005, T115, 1011−1014.(34) Rehr, J. J.; Albers, R. C.; Zabinsky, S. I. Phys. Rev. Lett. 1992, 69,3397−3400.(35) Ankudinov, A. L.; Ravel, B.; Rehr, J. J.; Conradson, S. D. Phys.Rev. B 1998, 58, 7565−7576.

Journal of the American Chemical Society Article

dx.doi.org/10.1021/ja3003765 | J. Am. Chem. Soc. 2012, 134, 9664−96719670

(36) Safonova, O. V.; Tromp, M.; Bokhoven, J. A.; van; Groot, F. M.F.; de; Evans, J.; Glatzel, P. J. Phys. Chem. B 2006, 110, 16162−16164.(37) Glatzel, P.; Singh, J.; Kvashnina, K. O.; vanBokhoven, J. A. J.Am. Chem. Soc. 2010, 132, 2555−2557.(38) Ankudinov, A. L.; Rehr, J. J.; Low, J.; Bare, S. R. Phys. Rev. Lett.2001, 86, 1642−1645.(39) Ankudinov, A. L.; Rehr, J. J.; Low, J. J.; Bare, S. R. J. SynchrotronRad 2001, 8, 578−580.(40) Mortensen, J. J.; Hansen, L. B.; Jacobsen, K. W. Phys. Rev. B2005, 71, 035109.(41) Enkovaara, J.; Rostgaard, C.; Mortensen, J. J.; Chen, J.; Dułak,M.; Ferrighi, L.; Gavnholt, J.; Glinsvad, C.; Haikola, V.; Hansen, H. A.;Kristoffersen, H. H.; Kuisma, M.; Larsen, A. H.; Lehtovaara, L.;Ljungberg, M.; Lopez-Acevedo, O.; Moses, P. G.; Ojanen, J.; Olsen,T.; Petzold, V.; Romero, N. A.; Stausholm-Møller, J.; Strange, M.;Tritsaris, G. A.; Vanin, M.; Walter, M.; Hammer, B.; Hakkinen, H.;Madsen, G. K. H.; Nieminen, R. M.; Nørskov, J. K.; Puska, M.;Rantala, T. T.; Schiøtz, J.; Thygesen, K. S.; Jacobsen, K. W. J. Phys.:Condens. Matter 2010, 22, 253202.(42) Bahn, S. R.; Jacobsen, K. W. Comput. Sci. Eng. 2002, 4, 56−66.(43) Hammer, B.; Hansen, L. B.; Nørskov, J. K. Phys. Rev. B 1999, 59,7413−7421.(44) Larsen, A. H.; Vanin, M.; Mortensen, J. J.; Thygesen, K. S.;Jacobsen, K. W. Phys. Rev. B 2009, 80, 195112.(45) Bauer, E.; van der Merwe, J. H. Phys. Rev. B 1986, 33, 3657−3671.(46) Hoster, H. E.; Alves, O. B.; Koper, M. T. M. ChemPhysChem2010, 11, 1518−1524.(47) Lima, F. H. B.; Zhang, J.; Shao, M. H.; Sasaki, K.; Vukmirovic,M. B.; Ticianelli, E. A.; Adzic, R. R. J. Phys. Chem. C 2007, 111, 404−410.(48) Horsley, J. A. J. Chem. Phys. 1982, 76, 1451−1458.(49) Mansour, A. N.; Cook, J. W.; Sayers, D. E. J. Phys. Chem. 1984,88, 2330−2334.(50) Hamalainen, K.; Siddons, D. P.; Hastings, J. B.; Berman, L. E.Phys. Rev. Lett. 1991, 67, 2850−2853.(51) Garcia-Araez, N.; Climent, V.; Herrero, E.; Feliu, J. M.;Lipkowski, J. Electrochim. Acta 2006, 51, 3787−3793.(52) Rossmeisl, J.; Nørskov, J. K.; Taylor, C. D.; Janik, M. J.;Neurock, M. J. Phys. Chem. B 2011, 110, 21833−21839.(53) Feibelman, P. J.; Esch, S.; Michely, T. Phys. Rev. Lett. 1996, 77,2257.(54) Abrams, B. L.; Vesborg, P. C. K.; Bonde, J. L.; Jaramillo, T. F.;Chorkendorff, I. J. Electrochem. Soc. 2009, 156, B273−B282.(55) Stamenkovic, V.; Mun, B. S.; Mayrhofer, K. J. J.; Ross, P. N.;Markovic, N. M.; Rossmeisl, J.; Greeley, J.; Nørskov, J. K. Angew.Chem., Int. Ed. 2006, 45, 2897−2901.

Journal of the American Chemical Society Article

dx.doi.org/10.1021/ja3003765 | J. Am. Chem. Soc. 2012, 134, 9664−96719671